{"id":{"repo_id":"reykjavik","oai_identifier":"oai:skemman.is:1946/29136"},"canonical_url":"https://search.dev.ndltd.org/etd/reykjavik/oai:skemman.is:1946/29136","repository":{"repo_id":"reykjavik","name":"Reykjavík University","base_url":"https://skemman.is/oai/request"},"display":{"title":"Geometric Verification of Image Copy Detection Results","abstract":"Many current image recognition methods are based on identifying interest points in images and considering images a match if they share similar interest points. This does not consider the locations of the interest points, so to reduce the number of false positives, geometric verification is used to identify and reject image pairs that do not fit into an expected geometric pattern. Designing a good geometric verification requires finding a balance between false positives and false negatives, while being time efficient. The current method used by Videntifier Technologies is based on comparing the angles between interest points. It works well but handles certain transformations incorrectly and is not very efficient. This report details an attempt to replace it with a method based on the RANSAC algorithm, which involves searching for an underlying transformation between the images that fits most of their matching interest points. After adding some improvements, using the RANSAC method resulted in fewer false negatives, without increasing the number of false positives or reducing time efficiency.","abstract_html":"Many current image recognition methods are based on identifying interest points in images and considering images a match if they share similar interest points. This does not consider the locations of the interest points, so to reduce the number of false positives, geometric verification is used to identify and reject image pairs that do not fit into an expected geometric pattern. Designing a good geometric verification requires finding a balance between false positives and false negatives, while being time efficient. The current method used by Videntifier Technologies is based on comparing the angles between interest points. It works well but handles certain transformations incorrectly and is not very efficient. This report details an attempt to replace it with a method based on the RANSAC algorithm, which involves searching for an underlying transformation between the images that fits most of their matching interest points. After adding some improvements, using the RANSAC method resulted in fewer false negatives, without increasing the number of false positives or reducing time efficiency.","abstract_has_math":false,"creators":["Stefanía Bergljót Stefánsdóttir 1994-"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Háskólinn í Reykjavík"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2017,"date_issued":"2017-10-05T14:08:16Z","date_published":"2017-10-05T14:08:16Z","updated_at":"2026-07-27T20:38:54Z","subjects":["Tölvunarstærðfræði","Discrete mathematics and computer science"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/1946/29136","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Háskólinn í Reykjavík"]},{"key":"dc:creator","label":"Author","values":["Stefanía Bergljót Stefánsdóttir 1994-"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2017-10-05T14:08:16Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2017-10-05T14:08:16Z"]},{"key":"dc:date.issued","label":"Date","values":["2017-10-05T14:08:16Z"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Tölvunarstærðfræði","Discrete mathematics and computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/1946/29136"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Many current image recognition methods are based on identifying interest points in images and considering images a match if they share similar interest points. This does not consider the locations of the interest points, so to reduce the number of false positives, geometric verification is used to identify and reject image pairs that do not fit into an expected geometric pattern. Designing a good geometric verification requires finding a balance between false positives and false negatives, while being time efficient. The current method used by Videntifier Technologies is based on comparing the angles between interest points. It works well but handles certain transformations incorrectly and is not very efficient. This report details an attempt to replace it with a method based on the RANSAC algorithm, which involves searching for an underlying transformation between the images that fits most of their matching interest points. After adding some improvements, using the RANSAC method resulted in fewer false negatives, without increasing the number of false positives or reducing time efficiency."]},{"key":"dc:title","label":"Title","values":["Geometric Verification of Image Copy Detection Results"]}]}],"canonical_facts":{"dc:contributor":["Háskólinn í Reykjavík"],"dc:creator":["Stefanía Bergljót Stefánsdóttir 1994-"],"dc:date.accessioned":["2017-10-05T14:08:16Z"],"dc:date.available":["2017-10-05T14:08:16Z"],"dc:date.issued":["2017-10-05T14:08:16Z"],"dc:description.abstract":["Many current image recognition methods are based on identifying interest points in images and considering images a match if they share similar interest points. This does not consider the locations of the interest points, so to reduce the number of false positives, geometric verification is used to identify and reject image pairs that do not fit into an expected geometric pattern. Designing a good geometric verification requires finding a balance between false positives and false negatives, while being time efficient. The current method used by Videntifier Technologies is based on comparing the angles between interest points. It works well but handles certain transformations incorrectly and is not very efficient. This report details an attempt to replace it with a method based on the RANSAC algorithm, which involves searching for an underlying transformation between the images that fits most of their matching interest points. After adding some improvements, using the RANSAC method resulted in fewer false negatives, without increasing the number of false positives or reducing time efficiency."],"dc:identifier.uri":["http://hdl.handle.net/1946/29136"],"dc:language.iso":["en"],"dc:subject":["Tölvunarstærðfræði","Discrete mathematics and computer science"],"dc:title":["Geometric Verification of Image Copy Detection Results"],"dc:type":["Thesis"]},"updated_at":"2026-07-27T20:38:54Z"}